AI-native commerce data platform

AI adoption stalls on trust.
We build the trust layer.

Dataloader loads every commerce system you run, resolves it into one source of truth, audits it, and enriches it — with every value tagged by where it came from. So your agents, feeds and dashboards act on data you can prove.

observatory://pipeline LIVE
LoadersCoreOutputs
ERP
Accounting
POS & stores
Warehouse / WMS
PIM
Shopify
Google Ads · Meta
Analytics · Klaviyo
The Observatory
resolve · audit · enrich · measure
Merchant Center feed
Schema.org JSON-LD
Shopify write-back
Agents · MCP
Insights
Business intelligence
12:33:12forecaststock cover projectedEstimated
12:33:14accountinginvoice lines matched to ordersObserved
12:33:16enrichdescription rewritten for searchGenerated
12:33:18adsPOAS computed per productDerived
12:33:20shopifyproduct.price ingestedObserved
›_
ERPAccountingPOSWMSPIMCRMShopifyGoogle AdsGoogle Merchant CenterSearch ConsoleGoogle AnalyticsMetaKlaviyoERPAccountingPOSWMSPIMCRMShopifyGoogle AdsGoogle Merchant CenterSearch ConsoleGoogle AnalyticsMetaKlaviyo
01 — The pipeline

Six stages. One source of truth.

From siloed exports to agent-ready data. Click a stage, or let it run.

01 · Load

Every source, field-scoped.

Observers connect to Shopify, Google Ads, Merchant Center, Search Console, Analytics, Meta and Klaviyo out of the box. Only the fields we need are requested, and raw history lands in The Archive — compressed and content-addressed.

$ observe shopify --scope catalog,orders
$ observe google-ads --scope campaigns
$ archive put --hash sha256 --zstd
$ ok · no PII requested
$ ▍
02 — Provenance & traceability

Every number traces back to its source.

Provenance tells you what kind of value you are looking at. Traceability shows you how it got there. Every value, metric and data gap carries both: the source it came from, every step it passed through, and the rule, formula or model that touched it.

Metrics trace to their exact inputs and formula.
Data traces to the system and the moment it was loaded.
Data gaps trace to the sources they are missing from — and how they were filled.
AI output traces to its grounding, so a fact never hides a guess.
Value states
ObservedDeclaredEstimatedDerivedGenerated
entity · product [EXAMPLE RECORD] · click a field to trace it
trace · margin 5 steps · source → value
Input price €129.00 from Shopify trace → price
Input cost €75.85 from ERP erp · purchase_price · 2026-11-15
Audit Rule passed: cost is below price check: cost < price
Measure margin = (price − cost) / price engine = sql · deterministic · llm = none
Value Same inputs always give the same result state = Derived
03 — What you get

From trusted data to growth — in your store, your ads and your decisions.

Every solution runs on the same traceable dataset, so what your customers see and what your team decides on always agree.

Metrics layer

Numbers you can trust

Every KPI is calculated deterministically with rules agreed with you. The same question always returns the same number — and every number traces back to its source.

marginPOASROASstock cover
Dashboards & insights

See what drives growth

Live dashboards and AI analyses compare your numbers and surface opportunities — where sales and profit really come from, and what to do next.

Content

AI product data

Complete descriptions, attributes and categories for thousands of products. The same data feeds product pages, Google Shopping, search engines and AI search.

Storefront

AI shopping assistant

A product advisor that lets shoppers describe what they need in their own words, compares options and guides them to the right product — and shows you what customers ask before they buy.

Storefront Coming soon

Semantic product search

Shoppers find the right product from everyday language, without knowing the exact name, model or technical term.

Commercial AI steering

Advertise on real profit

Combines sales, margin, stock and ad results to classify every product — separately for advertising profitability and stock productivity — so budgets follow your actual business, not just platform numbers.

ScaleKeepWatchFixDead
Automation

Agents & workflows

Agent fleets that take over the repetitive work your team does by hand today, running on the same trusted dataset.

Open access

BYOA: Bring your own AI

Connect Claude, ChatGPT or your own agents over MCP to ask questions and build analyses and dashboards straight on your data.

04 — Outcomes

What our customers achieved.

By organising all of their data and running on our systems. Each result is labelled with where the number comes from.

Growth · webshop 01
+0%

Sales vs last year — plus 0.5 pp higher site-wide conversion.

Last year
100
This year
140
Insights dashboards and insights made marketing more effective.
AI sales advisor live on 40,000+ products.
AI search shoppers find the right product faster.
Faster to market enrichment gets new products live sooner.
Enrichment labels across all data so AI recommends and analyses better.
MCP + fixed KPIs the team analyses its own data with AI, on numbers that always match.
Owner-reported customer result
Growth · omnichannel 02
+0%

Sales vs last year — 3 storefronts, 5 warehouses and POS in one order flow.

Last year
100
This year
200
Order flow orders routed automatically between storefronts, warehouses and POS.
Exceptions flagged for a person, everything else runs hands-free.
Inventory sync stock synced across all warehouses and storefronts.
Enrichment products tagged and organised for better categories and collections.
One catalogue the same enriched data published to every storefront.
Advisor AI product advisor answering from that same data.
Owner-reported customer result
Launch 03
0

Enriched for buyer intent. Their most successful launch ever.

At once: all 3,000 products enriched in one run.
GEO: structured so AI search engines can find and quote it.
Day one: sales right after launch. A first for this customer.
Owner-reported customer result
Warehouse 04
Up to 0×

Warehouse performance optimised — and stock placed where it sells.

Shippers: dispatch time per shipper — up to 3× apart, now visible.
Order status: orders in the warehouse tracked live, by status.
Placement: where to stock each product, based on its popularity.
Bought together: products that sell together stocked in the same warehouse.
Measured in Dataloader dispatch data, 2026
Marketing 05

Better marketing performance from product and campaign insights.

Products: performance per product and channel, on real margin.
Campaigns: which campaigns drive profit, not just revenue.
Gaps: products missing from channels, found and fixed.
Steering: budgets follow Scale, Keep, Watch, Fix and Dead.
Owner-reported customer result
Reporting 06

Easy reporting on sales, revenue and stock.

Sales & revenue: by product, channel and store, always up to date.
Stock: levels and cover across every warehouse.
Self-serve: ask in Claude or ChatGPT and get the report.
Same numbers: fixed KPIs, so every report agrees.
Owner-reported customer result
Found & fixed 07

Errors caught at the source. Dead stock turned into cash.

COGS: ERP and shop cost prices that did not match, flagged.
Site & feeds: zero prices, false negative stock and feed disapprovals fixed.
Tracking: GA4 issues found before they skewed numbers.
Dead stock: found by sales speed and margin, cleared for cash.
Owner-reported customer result
05 — Trust

European by design. Yours by default.

Your data stays yours. We provide a managed data layer hosted entirely in Europe, leaving you in full control of what you share and what you take back.

HOSTING Dedicated Helsinki infrastructure ISO/IEC 27001-certified Hetzner facilities. We keep no production copy of your data outside the EEA.
PRIVACY Minimise personal data Customer data is left out by design. We process no profiles, individual events, or click IDs.
AI Providers barred from training Models run in the EU or EU-hosted regions and receive only redacted, pseudonymised data, under contract.
ACCESS Read-only access grants We use read-only grants wherever allowed. Every grant is visible and removable in your own tools.
OWNERSHIP Export anytime, no charge Your data and anything derived from it stays yours. Full access via API, MCP, and exports.
EXIT Returned or deleted entirely On exit, data is returned or deleted. Live systems clear in 30 days, backups in 90.
Where is our data stored and processed?

Dedicated infrastructure in Helsinki, Finland.

Dedicated infrastructure in Helsinki, Finland (Hetzner, ISO/IEC 27001-certified facilities). No production copy outside the EEA. Secrets in an encrypted, self-hosted vault. Language models run with providers in the EU or EU-hosted regions, receive only redacted, pseudonymised data, and are contractually barred from training on your data. Generated values are tagged so they stay distinguishable from observed ones.

✓Dedicated infrastructure in Helsinki, Finland (Hetzner, ISO/IEC 27001-certified facilities). No production copy outside the EEA. Secrets in an encrypted, self-hosted vault.
✓Place of Growth Oy, trading as Dataloader. Finnish company, Finnish law.
✓GDPR Article 28 data processing agreement, security documentation and a security questionnaire on request. Hosting facilities are ISO/IEC 27001-certified (Hetzner); Dataloader itself makes no ISO or SOC 2 claim.
Do you collect personal customer data?

Minimise personal data by design.

Personal customer data is left out by design: no profiles, no individual events, no click ids.

✓Personal customer data is left out by design: no profiles, no individual events, no click ids.
Will our data be used to train language models?

Providers are contractually barred from training on your data.

Language models run with providers in the EU or EU-hosted regions, receive only redacted, pseudonymised data, and are contractually barred from training on your data. Generated values are tagged so they stay distinguishable from observed ones.

✓Language models run with providers in the EU or EU-hosted regions, receive only redacted, pseudonymised data, and are contractually barred from training on your data. Generated values are tagged so they stay distinguishable from observed ones.
How do we know which of our KPIs we can trust?

Every KPI gets a confidence rating from your connected data.

Available data determines KPI confidence. GMROI is gross margin return on inventory investment: gross margin for a period divided by average inventory at cost for that period. Missing purchase costs prevent a reliable calculation.

  • Evaluates 150 distinct KPIs across 11 business domains.
  • Assigns explicit confidence levels per metric: High, Reasonable, Low, Very low, No data.
  • Records the exact source system and ingest run for every value.
  • Flags missing inputs before they inform decisions.
✓We maintain a register covering 150 KPIs across 11 domains, rating each against the customer's connected data.
✓Every value carries its provenance: source, ingest run, time, and who produced it (a person, a process or an AI agent), and whether it was observed, declared, estimated, derived or generated.
How do you handle API access and system credentials?

Control access with read-only grants.

Read-only access wherever the provider allows it. Every grant is visible in your own tools and can be removed there.

✓Read-only access wherever the provider allows it. Every grant is visible in your own tools and can be removed there.
✓Reproducible infrastructure (NixOS), services in Rust, PostgreSQL, encryption in transit and at rest, weekly security reviews, no production data in development.
What happens if we decide to leave?

Retain full ownership of your data.

Your data and everything derived from it stays yours. Full access at any time through the API, the MCP server and exports, at no extra charge. On exit we return it in a machine-readable format or delete it (30 days for live systems, 90 for backups).

✓Your data and everything derived from it stays yours. Full access at any time through the API, the MCP server and exports, at no extra charge. On exit we return it in a machine-readable format or delete it (30 days for live systems, 90 for backups).
How will you notify us of breaches or supply chain changes?

Breach notice within 24 hours; sub-processor changes announced 30 days ahead.

Personal data breaches reported to you within 24 hours. Sub-processor changes announced 30 days ahead.

✓Personal data breaches reported to you within 24 hours. Sub-processor changes announced 30 days ahead.
KPI REGISTER
150 KPIs · 11 domains

Which of your KPIs can you actually trust? We maintain a register of KPIs across business domains. Before we show you a dashboard, we rate every KPI against the data we connected for you. Empty is better than wrong.

HighReasonableLowVery lowNo data
DOCUMENTS ON REQUEST
  • GDPR Article 28 data processing agreement (DPA)
  • Security documentation
  • Security questionnaire
Talk to us

See your business through the Observatory.

Standard sources connect out of the box. Anything else — ERP, accounting, HR — we model and map for you.

Book a demo Talk to us

Book a demo

See your business through the Observatory.

Tell us who you are and we'll get back to you to set up a demo.